Publication number | US20060002291 A1 |

Publication type | Application |

Application number | US 10/880,887 |

Publication date | Jan 5, 2006 |

Filing date | Jun 30, 2004 |

Priority date | Jun 30, 2004 |

Also published as | US7561534 |

Publication number | 10880887, 880887, US 2006/0002291 A1, US 2006/002291 A1, US 20060002291 A1, US 20060002291A1, US 2006002291 A1, US 2006002291A1, US-A1-20060002291, US-A1-2006002291, US2006/0002291A1, US2006/002291A1, US20060002291 A1, US20060002291A1, US2006002291 A1, US2006002291A1 |

Inventors | Mansoor Alicherry, Iraj Saniee, Sudipta Sengupta |

Original Assignee | Lucent Technologies, Inc. |

Export Citation | BiBTeX, EndNote, RefMan |

Patent Citations (10), Referenced by (25), Classifications (13), Legal Events (3) | |

External Links: USPTO, USPTO Assignment, Espacenet | |

US 20060002291 A1

Abstract

A number of techniques are described for routing methods that improve resistance to faults affecting groups of links subject to common risks. One of these techniques accounts for failure potentials in physical networks by considering shared risk link groups separately from performance and costs metrics in determining a primary routing path and a backup path. A shared risk link group (SRLG) is an attribute attached to a link to identify edges that have physical links in common and can therefore be simultaneously disrupted due to a single fault. Another technique considers node disjointness and provides a solution of two paths that are as node disjoint as possible and minimizes administrative costs. The techniques may further be combined in a priority order thereby providing a solution of at least two paths that are strictly SRLG disjoint, as node-disjoint as possible, and have minimum administrative costs. Due to the priority order of evaluation and typical network physical configurations of links, with the links associated common fault SRLGs, the priority ordering technique is very efficient in determining at least two paths for routing between a source and destination node.

Claims(20)

defining a graph G to represent the physical network with nodes of G representing connection nodes of the physical network and edges of G representing links of the physical network;

assigning shared risk of link failure costs to a plurality of the edges of graph G;

identifying the first primary path from a first node of the network to a second node of the network based on graph G with assigned shared risk of link failure costs; and

identifying the first backup path from the first node of the network to the second node of the network based on a reduced edge residual graph from which, at least one edge that shares a risk of failure with an edge of the first primary path, has been removed.

removing from the graph G each edge of the first primary path to define a first residual graph;

removing from the first residual graph each edge that shares a risk of failure with an edge of the first primary path to define a second residual graph; and

assigning node costs and edge costs to the second residual graph to define the reduced edge residual graph.

identifying a plurality of sets of edges, where each set has edges corresponding to links that share a common failure risk and where an edge may be a member of more than one set, the edges of each set having a shared risk link group (SRLG) attribute;

calculating the sum of the SRLG attributes in each set of edges as an SRLG size metric; and

assigning to each edge in the graph G the largest SRLG size that is associated with the edge.

constructing for each node on the first primary routing path a replacement node made up of multiple internal nodes that are interconnected by identifiable paths;

assigning a node cost to each identifiable path associated with incoming and outgoing paths not used on a reversal of the first primary routing path; and

assigning to one or more other identifiable paths a zero cost.

defining a network graph G to represent a physical network with nodes of G representing connection nodes of the physical network and edges of G representing links of the physical network;

identifying a network path from a starting node to an ending node in network graph G;

selecting nodes on the network path to be replacement nodes;

constructing **2** *b *internal nodes for each replacement node, where the value b is the number of incoming paths incident to the associated node in a reduced edge residual graph from which, at least one edge that shares a risk of failure with an edge of the network path, has been removed; and

interconnecting the **2** *b *internal nodes in each replacement node by identifiable paths, each identifiable path representing a valid path of the reduced edge residual graph.

removing the network path from the network graph G to create a residual graph; and

removing from the residual graph each edge that shares a risk of failure with an edge of the network path to define a reduced edge residual graph.

in each replacement node, assigning a node cost to identifiable paths, associated with incoming and outgoing paths not used on the reversal of the network path; and

in each replacement node, assigning a zero cost to one or more other identifiable paths.

defining a graph G to represent the physical network with nodes of G representing connection nodes of the physical network and edges of G representing links of the physical network;

assigning shared risk of link failure costs to a plurality of the edges of graph G;

identifying the first primary routing path from a first node of the network to a second node of the network based on graph G with assigned shared risk of link failure costs; and

identifying the first backup routing path from the first node of the network to the second node of the network based on a reduced edge residual graph from which, at least one edge that shares a risk of failure with an edge of the first primary routing path, has been removed.

removing from the graph G each edge of the first primary routing path to define a first residual graph;

removing from the first residual graph each edge that shares a risk of failure with an edge of the first primary routing path to define a second residual graph; and

assigning node costs and edge costs to the second residual graph to define the reduced edge residual graph.

identifying a plurality of sets of edges, where each set has edges corresponding to links that share a common failure risk and where an edge may be a member of more than one set, the edges of each set having a shared risk link group (SRLG) attribute;

calculating the sum of the SRLG attributes in each set of edges as an SRLG size metric; and

assigning to each edge in the graph G the largest SRLG size that is associated with the edge.

constructing for each node on the first primary routing path a replacement node made up of multiple internal nodes interconnected by identifiable paths;

selecting a node cost that is a largest administrative cost associated with paths in the physical network;

assigning the node cost to each identifiable path associated with incoming and outgoing paths not used on a reversal of the first primary routing path; and

assigning to one or more other identifiable paths a zero cost.

configuring the physical network to use the second primary routing path as the path selected for communicating information in the physical network; and

switching from the second primary routing path to the second backup routing path upon detection of a failure on the second primary routing path.

identifying a first path between first and second nodes in the graph, based on shared risk link group (SRLG) function costs associated with edges in the graph; and

determining a residual graph by removing edges from the graph representing the physical network, each removed edge sharing a risk of failure in common with an edge of the first path.

selecting nodes to be graphically represented in an expanded form; and

constructing for each selected node an expanded node made up of multiple internal nodes interconnected by identifiable paths associated with paths of the residual graph.

assigning a node cost to identifiable paths not used on the reversal of the first path;

assigning a zero node cost to one or more other identifiable paths;

identifying a second path in the residual graph between the first and second nodes in the physical network based on the assigned node costs.

assigning SRLG attributes to edges sharing a common risk of failure;

calculating for each edge a function of the SRLG attributes associated with each risk of failure; and

assigning to each edge in the graph the largest SRLG function cost that is associated with the edge.

Description

- [0001]The present invention relates generally to routing methods for determining paths in interconnection networks for communication or computer use.
- [0002]Communication networks take many forms for interconnecting a source node to a destination node. With the advent of the Internet and its exponential growth, communication networks are increasing in size and complexity and moving toward using optical networks to provide very high speed interconnection bandwidth of the order of gigabits to terabits per second. An optical network generally consists of optical switch nodes and fiber optic interconnection links between nodes arranged in a general mesh topology. Due to the size and complexity of many communication networks, routing paths between numerous source and destination nodes that meet various constraints is a difficult problem. Some of the constraints considered, for example, are performance, in terms of path distance and bandwidth, costs, such as fiber and equipment costs, and link and node failure probabilities.
- [0003]A physical network is typically modeled by converting point to point links to edges in a directed or undirected graph with nodes in the graph corresponding to optical switches and source and destination points of the physical system. The nodes and edges may be labeled with a metric, typically performance and costs, as a constraining parameter on the link or node. The problem of finding routing paths in a physical network corresponds to finding paths in the directed or undirected graph. Failure probabilities have been typically accounted for by routing two paths, a primary path and a backup path, and making both paths as disjoint as possible.
- [0004]One approach to finding a node/edge disjoint pair of paths between a given node pair in a directed or undirected graph is to use a technique such as described by Suurballe, “Disjoint paths in a Network,” Networks, Vol. 4, pp. 125-145, 1974. Since many characteristics of physical networks are not easily abstracted to a graph, using techniques such as described by Suurballe are not adequate, especially when considering failure potentials in physical networks.
- [0005]Among its many aspects, one embodiment of the present invention addresses a method of determining a first primary routing path and a first backup routing path in a physical network. A graph G is defined to represent the physical network with nodes of G representing connection nodes of the physical network and edges of G representing links of the physical network. Shared risk of link failure costs are assigned to a plurality of edges of graph G. A first primary routing path from a first node of the network to a second node of the network is then identified based on graph G with assigned shared risk of failure costs. A first backup routing path form the first node of the network to a second node of the network is then identified based on a reduced edge residual graph from which, at least one edge that shares a risk of failure with an edge of the first primary routing path, has been removed.
- [0006]Also, among its several aspects, another embodiment of the present invention addresses a method for constructing replacement nodes in a network graph. The network graph G being defined to represent a physical network with nodes of G representing connection nodes of the physical network and edges of G representing links of the physical network. A network path from a starting node to an ending node is identified and nodes on the network path are selected to be replacement nodes. For each replacement node,
**2***b*internal nodes are constructed, where the value b being the number of incoming paths incident to the associated node in a reduced edge residual graph from which, at least one edge that shares a risk of failure with an edge of the network path, has been removed. The**2***b*internal nodes in each constructed node are then interconnected by identifiable paths that represent valid network paths of the reduced edge residual graph where, the valid paths connect through the replacement node. - [0007]A more complete understanding of the present invention, as well as other features and advantages of the invention, will be apparent from the following detailed description and the accompanying drawings.
- [0008]
FIG. 1 illustrates an exemplary optical mesh network; - [0009]
FIG. 2A illustrates an undirected sub graph of the optical mesh network ofFIG. 1 , to be used in discussing steps of a prior art technique, illustrating the connection links and nodes between a node**1**and a node**6**; - [0010]
FIG. 2B illustrates a shortest path between node**1**and node**6**to be used to discuss steps of a prior art technique; - [0011]
FIG. 2C illustrates a residual graph determined using the shortest path found inFIG. 2B to be used to discuss steps of a prior art technique; - [0012]
FIG. 2D illustrates a shortest path found on the residual graph ofFIG. 2C to be used to discuss steps of a prior art technique; - [0013]
FIG. 2E illustrates one aspect of a detangling step with a merger of the two shortest paths found inFIG. 2B andFIG. 2D and highlights an example of redundant edges to be used to discuss steps of a prior art technique; - [0014]
FIG. 2F illustrates another aspect of a detangling step where two disjoint paths are obtained after the removal of the redundant edges ofFIG. 2E to be used to discuss steps of a prior art technique. - [0015]
FIG. 3A illustrates one typical scenario encountered in a portion of a physical network where communication links are physically routed together to pass through a conduit causing a shared common risk of failure; - [0016]
FIG. 3B illustrates a graph abstraction of the portion of the physical network ofFIG. 3A where physical aspects of routing through a conduit are not depicted; - [0017]
FIG. 4A illustrates a typical scenario encountered in a portion of a physical network where communication links are physically routed together to pass through a bridge or a tunnel causing a shared common risk of failure; - [0018]
FIG. 4B illustrates a graph abstraction of the portion of the physical network ofFIG. 4A where physical aspects of routing through a bridge or a tunnel is not depicted; - [0019]
FIG. 5 illustrates a presently preferred second program flow chart of an improved SRLG disjoint path routing process in accordance with the present invention; - [0020]
FIG. 6 illustrates a presently preferred third program flow chart of an SRLG disjoint path routing process that determines strict SRLG disjoint paths that are as node disjoint as possible, have minimum path length, and have minimal administrative costs in accordance with the present invention; - [0021]
FIG. 7A illustrates a residual graph similar to the graph ofFIG. 2C with an additional link; and - [0022]
FIG. 7B illustrates an advantageously constructed node graph for a node on the primary lowest cost path used to assign node costs appropriately dependent upon how a node is traversed in specific directions, in accordance with the present invention. - [0023]
FIG. 1 shows a physical network of connection nodes connected by links. InFIG. 1 , the physical network is embodied as an optical network made up of optical switch nodes connected by optic fiber links in a general mesh topology. - [0024]Links
**104**and link**120**are considered to be standard interface connections, which may be optical connections, to access node devices A**102**and B**122**, respectively. The links, such as**108**,**112**,**116**,**124**,**128**,**132**, and**136**, are considered to use optical fibers that may carry multiple optical channels. The nodes**106**,**110**,**114**,**118**,**126**, and**130**are optical switch nodes that can switch an incoming light path to an outgoing light path. Since optical switches are designed to be redundant, the typical type of failure that occurs in optical networks is damage to the fiber optic lines. - [0025]A particular example of a routing problem is the problem of finding a disjoint pair of paths in
FIG. 1 between node**1****106**and node**6****118**. The physical network, represented by the outlined area**140**, is modeled on an undirected graph**200**ofFIG. 2A , by converting point to point fiber links to edges in an undirected graph with nodes in the graph corresponding to the optical switches of the physical system. Specifically, links**108**,**112**,**116**,**124**,**128**,**132**, and**136**ofFIG. 1 correspond to edges**208**,**212**,**216**,**224**,**228**,**232**, and**236**ofFIG. 2A , respectively. Since each link and corresponding edge represent bidirectional communication paths, graph**200**is considered an undirected graph. Nodes**106**,**110**,**114**,**118**,**126**, and**130**ofFIG. 1 correspond to nodes**206**,**210**,**214**,**218**,**226**, and**230**ofFIG. 2A , respectively. - [0026]A well known technique to find a node/edge disjoint pair of paths between a given node pair in a directed or undirected graph is described by J. W. Suurballe, “Disjoint paths in a network”,
*Networks*, vol. 4, pp. 125-145, 1974. The technique requires solving a shortest path problem twice: once from a source node to a destination node using a directed or undirected graph, and the second time on a residual graph, again from the source node to the destination node. The residual graph is a subgraph of the original graph that remains after directed edges on the first shortest path, in the direction of the source to destination, have been sequentially removed. In an undirected graph, each edge can be thought of as having two edges that are directed but in opposite directions. The solution for the shortest path of the residual graph, if it exists, may require a post-processing step to remove edges that may be traversed in opposite directions in the pair of shortest paths selected. - [0027]
FIGS. 2A-2F show a typical evolution of applying a prior art technique to undirected graph**200**ofFIG. 2A to find a pair of disjoint paths between node**1****206**and node**6****218**. First, a shortest path is found in graph**240**outlined by box**244**ofFIG. 2B . Second, directed edges on the shortest path are removed, which leaves the reverse path from node**6**to node**1**as outlined in box**254**ofFIG. 2C , to create the residual graph**250**. Third, the shortest path on the residual graph**250**is found as indicated by dotted line**264**of graph**260**ofFIG. 2D . Fourth, the two discovered shortest paths are merged in graph**270**ofFIG. 2E thereby creating two redundant edges between node**3**and node**4**as highlighted by common edges**274**. Fifth, a post-processing step is performed to remove shared redundant edges, the common edges**274**ofFIG. 2E , leaving graph**280**with two disjoint paths as indicated by dotted lines**284**and**288**. Such merging of paths and elimination of edges is called detangling. - [0028]Though the technique described above with reference to
FIGS. 2A-2F is efficient, the idea of diversity as applied to a graph, that is, in the sense of node or edge disjointness, may not be adequate for diversity in a physical network. In a communication network, edges may share physical attributes not directly observable when the network is abstracted by a graph. Shared risk link groups (SRLGs) are a mechanism to identify groups of edges that have physical segments in common and which can therefore be simultaneously disrupted due to a single fault. Shared risk of link failures may arise most commonly in two scenarios as shown inFIGS. 3A and B andFIGS. 4A and B. - [0029]To track shared risks, each edge in a network graph is assigned one or more attributes, typically in an integer form, that represent various shared risks due to physical attributes of the network. For example,
FIG. 3A illustrates a network sub-section**300**of a larger physical network with links**304**,**308**, and**312**that, due to physical placement, are all routed through a conduit**315**. For example, such physical routing may occur when multiple communication links requiring connection to a central server are physically placed in the same conduit in a wall of a building. InFIG. 3B , a network graph abstraction**320**of the network sub-section ofFIG. 3A is shown, with corresponding edges**324**,**328**and**332**, which only abstracts connected links and nodes and consequently does not illustrate the confinement of the physical fibers in conduit**315**. In network sub-section**320**ofFIG. 3B , all three edges,**324**,**328**and**332**, belong to one SRLG since the three physical fiber links**304**,**308**, and**312**, corresponding to the three edges,**324**,**328**, and**332**, respectively, are all routed through a conduit**315**and consequently a single fault can affect all three links. - [0030]In another example,
FIG. 4A illustrates another network sub-section**400**of a larger physical network with links**404**and**408**that are physically placed through a tunnel, a bridge or the like,**412**. InFIG. 4B a network graph abstraction**420**of the network subsection ofFIG. 4A is shown, with corresponding edges**424**and**428**, which only abstracts connected links and nodes and consequently does not illustrate the physical fiber placement. InFIG. 4B , the parallel edges**424**and**428**belong to the same SRLG since a single fault can affect both links. In bothFIGS. 3B and 4B , edges in a shared risk link group are distinct and may even be disjoint as shown inFIG. 4B . Thus the graph abstraction of a communication network can fail to represent shared risks. - [0031]To formalize SRLGs, a network graph can be described by a three tuple (N, E, R) representing a physical network where N is a set of nodes, E is a set of edges in the network, and R is a set of elements, each of which consists of edges that share a common risk. Formally, R
_{i }ε R for 1≦i≦S, where each R_{i }is a set of edges in E and S depends on the network size. Thus, two edges e_{1 }and e_{2 }share a common risk if and only if e_{1}εR_{i }AND e_{2}εR_{i }for some R_{i}εR. We note that an edge may belong to multiple SRLGs. - [0032]Disjoint paths in a communication network can now be defined not only with respect to nodes and edges, but also, more accurately, relative to R. Thus, to improve a network's resistance to faults, a pair of paths is searched for between a node pair that are node/edge disjoint and have no risk groups in common. Unfortunately, this problem, in its full generality, is a hard problem, considered non-deterministic polynomial-time hard (NP-hard), primarily on account of shared risks of a type corresponding to those illustrated by way of example in
FIG. 4A . - [0033]Since a general physical network consists of many types of fault situations, including both of the types represented in
FIGS. 3A and 4A , it is difficult to establish an efficient, general and exact process for resolving an SRLG-disjoint path problem to cover a wide range of physical routing risks. Consequently, a number of specific heuristics are used and presented herein. - [0034]Among its many aspects, the present invention may be advantageously utilized to account for failure potential in physical networks, node disjointness of routing paths, minimum path length, and minimum administrative costs in efficiently finding primary and backup routing paths. Failure potentials in physical networks are accounted for by considering paths having common shared risks of failure and using a shared risk link group metric which is advantageously considered separately from performance and costs metrics in determining primary and backup paths. A shared risk link group (SRLG) is an attribute attached to a link to identify edges that have physical elements in common and can therefore be simultaneously disrupted due to a single fault. Node disjointness may be accounted for by an advantageous method of assigning node costs to minimize node overlaps in primary and backup paths. Path length and administrative costs may be minimized by advantageously assigning path length and node costs in determining a backup path after a primary path is found. Further, the present invention may advantageously evaluate these aspects of path failure potential, node disjointness, minimum path length, and administrative costs in a priority order thereby providing a solution of at least two paths that are strictly SRLG disjoint, as node-disjoint as possible, and have minimum administrative costs including minimum total path length. Due to the priority order of evaluation and typical network physical configurations of links, where the links are associated with common fault SRLGs, the priority ordering and cost metric assigning techniques are very efficient in determining at least two paths for routing between a source and destination node. To such ends, exemplary processes are described in detail below for SRLG disjoint routing.
- [0035]In a first process
**500**, the SRLGs are represented by R_{1}, R_{2}, . . . , R_{S }and a new variable |R_{i}| is introduced to denote the size of a shared risk link group. As an example, |R_{i}| may represent the number of edges that are in R_{i}, being a sum of the members of R_{i}. Process**500**illustrated inFIG. 5 uses size of SRLGs for the computation of the first path. - [0036]Process
**500**begins with defining a network graph G equal to (N, E, R) representing a physical network to find a primary routing path and a backup routing path between node pair (A, Z), initial step**504**. In a presently preferred embodiment, the primary routing path and the backup routing path are to be disjoint with respect to shared risk of link failures. The process continues by using SRLG sizes as a cost of each edge, that is C_{e}=|R_{i}| where |R_{i}| is the largest risk group edge e is a member of, where e ε E, step**508**. It is realized that a plurality of edges may be assigned SRLG costs due to practical considerations with zero or estimated SRLG costs assigned as appropriate. Next, K lowest cost paths p_{1}, p_{2}, . . . , p_{K }are computed using costs C_{e}, step**512**. Note that open shortest path first (OSPF) is a standard process that can be used to determine the lowest cost paths. This process determines a path based on a metric, such as assigned costs, and thus computes lowest cost paths even though it is called open shortest path first. Then, i is initialized to 1 in step**516**and the current lowest cost path is set to p_{i}, step**520**. Continuing with step**524**, the residual graph G_{i }of G is computed by removing from G, edges in p_{i }in the direction from A to Z. In the next step**528**, edges, in both directions, are removed from the residual graph G_{i }to eliminate edges that belong to any risk groups that edges in p_{i }are members of, excluding reversed links on path p_{i}, thereby creating a reduced edge residual graph G_{i}′. An administrative cost function H including path length is assigned in the next step**530**for each edge e, where e ε E. It is noted, that process**500**can use arbitrary administrative costs, where an administrative cost is any non-negative assignment of costs to edges of a graph. In process**500**, arbitrary administrative costs can be included through the cost function H in step**530**. The lowest cost path process that was run previously in step**512**is now run on the reduced edge residual graph G_{i}′ to find a lowest cost path q_{i }from A to Z using cost function H, step**532**. If a lowest cost path q_{i }was found, step**536**, then the process continues by detangling paths p_{i }and q_{i }and saves p_{i }and q_{i }or their detangled variants as the detangled paths p_{i}′ and q_{i}′, save step**540**. Process**500**then proceeds to step**544**. If a lowest cost path q_{i }was not found for this graph G_{i}, the process proceeds to step**544**. In step**544**it is determined if all paths for i≦K have been evaluated. If i is less than K, the process proceeds to increment i by one, step**548**, and then proceeds back to step**520**to repeat the process steps**520**to**536**to continue to try to find a disjoint pair of paths. If i is equal to K, then the process proceeds to end step**556**. The p_{i}′ and q_{i}′ with lowest administrative costs are selected from the saved results file or it is noted that no paths were found. - [0037]Although process
**500**addresses the requirement of SRLG disjointness and minimal length using the cost function H in step**530**as an administrative cost, it pays no attention to node disjointness. It may be desirable to have strict SRLG disjoint paths, which are also node disjoint and have small administrative costs. Unfortunately, it may not be possible to create node disjoint paths between a source node and a destination node. This may be due to a network, for example, that is designed to have single hub nodes serving end nodes. An example of such a network is network**100**ofFIG. 1 which has two such single hub nodes, node**1****106**and node**6****118**. Two disjoint routing paths between node**1****106**and node**6****118**can be determined as previously described above. However, in order to find paths between node A**102**and node B**122**, a path**104**from node A**102**to node**1****106**must be followed and to reach node B**122**a path**120**from node**6****118**must be followed. Consequently, two routing paths between node A**102**and node B,**122**, cannot be completely node disjoint due to the network construction and node**1****106**and node**6****118**must be common nodes on the two otherwise disjoint paths between node A**102**and node B**122**. - [0038]Thus, for a reasonable solution, it is important to have a strict SRLG-disjoint pair of paths that are as node-disjoint as possible and have minimal length and administrative costs. If these three conditions are interpreted in a priority order, a meaningful problem solution emerges. Without this priority, comparisons arise that may be hard to resolve. For example, consider a comparison between a pair of SRLG-disjoint paths with one node in common and total administrative cost of
**100**and a pair of SRLG-disjoint paths with two nodes in common and total administrative cost of**60**. With the priority ordering of the three requirements, a process, as described further below, can unambiguously select the first situation. - [0039]A second process
**600**, illustrated inFIG. 6 and described below, applies such priority ordering and a unique method of assigning node and edge costs to find routing paths with improved resistance to faults affecting groups of links subject to common risks. Process**600**begins by defining a network graph G equal to (N, E, R) representing a physical network to find a primary routing path and a backup routing path between node pair (A, Z), initial step**604**. As above, these two paths are preferably disjoint with respect to shared risk of link failures. The process continues by using SRLG sizes as a cost of each edge, that is C_{e}=|R_{i}| where |R_{i}| is the largest risk group edge e is a member of, where e ε E, step**608**. It is realized that a plurality of edges may be assigned SRLG costs due to practical considerations with zero or estimated SRLG costs assigned as appropriate. Next, K lowest cost paths p_{1}, p_{2}, . . . , p_{K }are computed using costs C_{e}, step**612**. Note that open shortest path first (OSPF) is a standard process that can be used to determine the primary routing path and backup routing path as lowest cost paths. This process determines a path based on a metric, such as assigned costs, and thus computes lowest cost paths even though it is called open shortest path first. Then, i is initialized to**1**in step**616**and the current lowest cost path is set to p_{i}, step**620**. Continuing with step**624**, the residual graph G_{i }of G is computed by removing from G edges on p_{i }in the direction from A to Z. In the next step**628**, remove from the residual graph G_{i }the edges, in both directions, that belong to any risk groups that edges in p_{i }are members of, excluding reversed links on path p_{i}, creating a reduced edge residual graph G_{i}′. Next, assign a fixed cost F equal to (N*max_admin)+1 to each node in p_{i }dependent upon selective paths associated with each node, where N is the number of nodes and max_admin is the largest administrative cost, step**632**as described in further detail below. This fixed cost is not charged if a node on p_{i }is entered from a directed link and exited from another directed link both of which belong to the set of remaining (reversed) links on p_{i}. By placing large costs on nodes on the primary path p_{i}, paths that are maximally node disjoint can be found. - [0040]Due to the detangling of the primary and backup paths computed via the residual graph, nodes traversed in opposite directions will not appear in the final detangled paths. Thus, such nodes, though they may be initially common between primary and backup paths, will not exist on the final detangled paths. These nodes therefore need not incur fixed costs.
- [0041]Among its several aspects, the present invention provides an advantageous approach for assigning node costs only if a node is traversed in specific directions. Using residual graph
**700**ofFIG. 7A , an exemplary node construct**750**is shown inFIG. 7B which illustrates an advantageously constructed node graph for a node on the primary lowest cost path used to assign node costs appropriately which is dependent upon how a node is traversed in specific directions. The nodes in both the residual graph**700**and node construct**750**use a unique letter symbol to represent each node with a path between nodes indicated by use of the → symbol. Residual graph**700**ofFIG. 7A shows a reversed primary flow**702**, C→A→D→E, after removal of a lowest cost primary flow, E→D→A→C. Node A**704**lies on the primary lowest cost flow computed in the first phase of processing, such as followed in the processing of graph**200**ofFIG. 2A producing residual graph**250**including a reversed primary flow**254**. The residual graph**700**has an additional bidirectional path**706**connecting node A**704**and node F**708**. - [0042]In considering node cost assignment, node A
**704**should incur no cost if traversed on a flow that includes an edge of the reversed primary flow, such as paths C→A→D, C→A→B, and B→A→D. However, paths F→A→B and B→A→F should incur fixed costs, as these flows will not result in detangling of node A**704**once combined with primary flow E→D→A→C. To obtain the desired end result, each node on the primary path, such as path**244**ofFIG. 2B , is replaced by**2***a*internal nodes, where a is the number of arcs into the node based on the residual graph. An undirected arc is equivalent to two arcs, one in each direction. - [0043]Thus, node
**4****704**with three arcs into it, C→A, B→A, and F→A, is replaced by a replacement node construct**752**made up of six internal nodes, Aa**754**, Ab**756**, Ac**758**, Ad**760**, Ae**762**, and Af**764**, as shown inFIG. 7B . Node F**708**, node D**710**, node B**712**, and node C**714**interface with node A**704**with either unidirectional edges, such as A→D edge**720**and C→A edge**724**, or bidirectional edges, such as {A→B and B→A} edge**728**and {A→F and F→A} edge**706**. The bidirectional edges are split into separate paths as shown inFIG. 7B such that node B**712**corresponds to the two nodes, node B**766**and node B′**767**, and node F**708**corresponds to two nodes, node F**768**and node F′**769**. Node D**710**corresponds to node D**770**and node C**714**corresponds to node C**772**. The bidirectional edge**728**corresponds to directed edge, Ae→B,**774**and directed edge, B′→Ad**775**. The bidirectional edge**706**corresponds to directed edge, Ac→F,**776**and directed edge, F′→Af,**777**. Directed edge**720**corresponds to directed edge, Aa→D,**778**and directed edge**724**corresponds to directed edge, C→Ab,**780**. - [0044]Each flow, such as C→A→B, in
FIG. 7A is now identifiable with a unique flow on the new replacement node construct**752**, such as C→Ab→Ae→B. The replacement node construct**752**shown inFIG. 7B consists of six internal nodes and new edges connecting these internal nodes, where each edge represents a valid connection path through node A**704**. Since there are seven ways to traverse node A**704**, C→A→D, C→A→B, C→A→F, B→A→D, B→A→F, F→A→D, and F→A→B, there are seven edges shown in dotted lines inFIG. 7B . Note that paths, such as path B′→Ad→Ae→B, have been excluded since, in this example, it makes little sense to fold on a link for cost routing purposes. A node cost is translated to replacement node internal edge costs. The cost on these edges is determined in step**632**ofFIG. 6 . For each node on p_{i}, for example node**1**, D, A, and C ofFIG. 7A , construct a replacement path identifiable node made up of**2***a*internal nodes, for example internal nodes Aa, Ab, Ac, Ad, Ae, and Af ofFIG. 7B , interconnected by identifiable paths, such as paths**784**,**788**, and**790**, and assign a cost F, equal to (N*max_admin)+1, to each identifiable path associated with incoming and outgoing paths not used on the reversed primary path p_{i}, such as Af→Ae path**784**and Ad→Ac path**788**, with other identifiable paths being assigned a zero cost, such as paths**790**, where a is the number of incoming paths incident to a node, N is the number of nodes, and max_admin is the largest administrative cost - [0045]Clearly, with a higher connectivity, a larger number of nodes and edges have to be added to the original residual graph, such as residual graph
**700**FIG. 7A . The new residual graph remains computationally tractable since at most max_degree*num_primary_path_nodes replacement internal nodes are used in modifying an original residual graph. Based on a residual graph, the max_degree variable is the maximum a, considering all the primary path nodes, where a is the number of incoming paths incident on a node. In addition, at most max_degree*max_degree new edges are added to the residual graph. Thus, the lowest cost path computations remain polynomial. In real networks, since the max_degree is around four to five, this construct does not add much to the complexity of finding a solution or to the computational times. - [0046]In the next step
**634**, an administrative cost function H is assigned for each edge e external to the constructed replacement nodes and e ε E. It is noted, that process**600**can use arbitrary administrative costs for edge cost function H, where administrative cost is any non-negative assignment of costs to edges of a graph. For example, administrative cost can include a number representing the inverse of a link's bandwidth such that higher bandwidth links are rewarded with lower costs to encourage their use. Another example of administrative costs can include a number representing latency delay in order to discourage the use of high latency links. - [0047]The lowest cost path process that was run previously in step
**612**is now run on the reduced edge residual graph G_{i}′ to find a lowest cost path q_{i }from A to Z using the assigned node and edge costs, step**636**. If a lowest cost path q_{i }was found, step**640**, the process detangles paths p_{i }and q_{i }and saves p_{i }and q_{i }or their detangled variants as the detangled paths p_{i}′ and q_{i}′ in a results file, step**644**. The two lowest cost paths p_{i}′ and q_{i}′ that have been found are SRLG disjoint, have low administrative cost and are maximally node disjoint. Process**600**then proceeds to step**648**where it is determined whether all K lowest cost paths have been evaluated. If all paths have not been evaluated, then the process proceeds to increment i by one, step**660**, and proceeds back to step**620**to repeat the steps**620**to**648**to continue to try to find a disjoint pair of paths. Back at step**640**, if a lowest cost path q_{i }was not found for this graph G_{i}′, then the process, for example, saves the fact that no q_{i }was found in a results file, step**656**, and proceeds to step**648**. If all paths have been evaluated, then the process proceeds to step**652**where the p_{i}′ and q_{i}′ with lowest administrative costs are selected from the saved results file or it is noted that no paths were found and process**600**stops. - [0048]Table 1 below illustrates the performance of process
**600**on two sets of realistic networks. Network**1**consists of 58 nodes and 101 links while network**2**consists of 24 nodes and 39 links. SRLG-disjoint paths were computed for all node pairs in each network and the average run time, the variance, and the degree of node-disjointness possible were measured. These simulation calculations were conducted on a Sun4u Sparc Ultra-Enterprise machine.TABLE 1 Value of K in K- Run Time (msec) Number of Node lowest cost paths Min/Mean/Max − Overlaps giving best solution, standard deviation Min/Mean/Max Min/Mean/Max Network 1 150/337/640 − 76 0/0/0 − 0 1/1.94/32 − 3.11 58 nodes, 101 links Network 2 10/155/220 − 33 0/0.37/2 − 0.72 1/1.33/7 − 0.85 24 nodes, 39 links - [0049]It is observed that in network
**1**node-disjoint paths were allowed between all node pairs and process**600**obtained all such primary and backup paths while network**2**had some node pairs for which node-disjoint paths were not possible. In all cases examined, the paths were also of minimal total administrative costs meeting one of the goals of the routing process for physical networks. - [0050]In use on physical networks, the two paths generated, for example, by process
**600**for a physical network are used to configure the physical network node switches to support one of the paths as a primary routing path and support the other diversely routed path as a backup path. A transmission is thereby set up to use the primary routing path unless a failure of the primary routing path is detected. Upon detection of a failure in the primary routing path, the physical network nodes switch to the backup routing path. For example, set up messages may be used to reconfigure the physical network nodes to the backup routing path. By having SRLG disjoint paths that are as node disjoint as possible and have minimum administrative costs, the switch to the backup routing paths will have a high degree of success. - [0051]Process
**600**has been developed using C/C++ code and can be run on standard personal computers such as computer system**800**shown inFIG. 8 . A computer system**800**, or the like, may suitably comprise a processor complex**804**, a user interface or interfaces such as a keyboard and mouse**808**, a display**812**, and output, such as printer**816**. Memory and hard drive**820**for storing and running programs such as C/C++ or other suitable code for implementing process**600**, and media interfaces**824**consisting of a diskette drive, cd drive, internet interface and the like, for transferring data and programs, such as process**600**with network graphs stored and manipulated as abstract data types, are also part of the computer system**800**. - [0052]Using a computer system, such as computer system
**800**, process**600**ofFIG. 6 has shown a faster rate of convergence to a solution than process**500**ofFIG. 5 . Realistic network convergence to solutions having run times on the order of 100 msecs have been achieved. Consequently, process**600**can be used in physical networks to statically determine routing paths or to dynamically route paths based upon real time network characteristics. - [0053]While the present invention has been disclosed in a presently preferred context, it will be recognized that the present teachings may be adapted to a variety of contexts consistent with this disclosure and the claims that follow. By way of example, while a presently preferred embodiment of the present invention is to find a strict SRLG-disjoint pair of routing paths, it is realized that practical or physical considerations may intervene. For example, there may be physical networks where the process of forming the reduced edge residual graph, such as described in step
**528**and in step**628**, is adjusted such that it may not remove all edges, in both directions, that belong to any risk groups that edges on p_{i }are members of. As an alternative for the edges not removed, a high administrative cost may be applied. In another example, while the present invention has been disclosed primarily in the context of optical networks, it will be recognized that the invention can be applied to routing processes for other network implementations, such as wired networks, and further can be used in planning stages of network deployment to improve a network's resistance to faults in the links between nodes in the network.

Patent Citations

Cited Patent | Filing date | Publication date | Applicant | Title |
---|---|---|---|---|

US5210837 * | Jun 15, 1990 | May 11, 1993 | Digital Equipment Corporation | Methods and apparatus for transforming machine language program control into high-level language constructs by manipulating graphical program representations |

US6678876 * | Aug 24, 2001 | Jan 13, 2004 | Formfactor, Inc. | Process and apparatus for finding paths through a routing space |

US6898184 * | Nov 15, 1998 | May 24, 2005 | Hewlett-Packard Development Company, L.P. | Private arbitrated loop self-test management for a fibre channel storage enclosure |

US20020105904 * | Jun 28, 2001 | Aug 8, 2002 | Oded Hauser | Maintaining information to optimize restorable dynamic routing with shared backup |

US20040218525 * | Apr 29, 2003 | Nov 4, 2004 | David Elie-Dit-Cosaque | Protection scheme for a communications network under multiple failures |

US20050031339 * | Jul 9, 2004 | Feb 10, 2005 | Chumning Qiao | Efficient trap avoidance and shared protection method in survivable networks with shared risk link groups and a survivable network |

US20050073958 * | Nov 19, 2004 | Apr 7, 2005 | Avici Systems, Inc. | Selecting alternate paths for network destinations |

US20050111349 * | Nov 21, 2003 | May 26, 2005 | Vasseur Jean P. | Method and apparatus for determining network routing information based on shared risk link group information |

US20050237950 * | Apr 25, 2005 | Oct 27, 2005 | Board Of Regents, The University Of Texas System | System, method and apparatus for dynamic path protection in networks |

US20050243711 * | May 3, 2004 | Nov 3, 2005 | Alicherry Mansoor A K | Method and apparatus for pre-provisioning networks to support fast restoration with minimum overbuild |

Referenced by

Citing Patent | Filing date | Publication date | Applicant | Title |
---|---|---|---|---|

US7697505 * | Apr 13, 2010 | Cisco Technology, Inc. | Techniques for decreasing queries to discover routes in an interior gateway protocol | |

US7944856 * | Dec 23, 2008 | May 17, 2011 | At&T Intellectual Property I, L.P. | Method and apparatus for providing redundancy for an access network |

US7986617 | Jul 26, 2011 | Noblis, Inc. | System and method for providing network resiliency | |

US8472347 | May 26, 2011 | Jun 25, 2013 | Noblis, Inc. | System and method for providing network resiliency |

US8488963 * | Jun 30, 2008 | Jul 16, 2013 | Infinera Corporation | Link diversity and load balancing across digital and optical express-thru nodes |

US8831012 * | Oct 7, 2009 | Sep 9, 2014 | Fujitsu Limited | Determining paths in a network with asymmetric switches |

US9112775 | Jun 24, 2013 | Aug 18, 2015 | Huawei Technologies Co., Ltd. | Method and device for establishing backup path, method and device for selecting backup path |

US9350481 * | Oct 7, 2013 | May 24, 2016 | Futurewei Technologies, Inc. | Transport functions virtualization for wavelength division multiplexing (WDM)-based optical networks |

US9350641 | Mar 30, 2012 | May 24, 2016 | Alcatel Lucent | IP fast reroute scheme offering full protection |

US20060173664 * | Jan 31, 2005 | Aug 3, 2006 | Suban Krishnamoorthy | SAN modeling |

US20070183334 * | Feb 3, 2006 | Aug 9, 2007 | Cisco Technology, Inc. | Techniques for decreasing queries to discover routes in an interior gateway protocol |

US20090324222 * | Jun 30, 2008 | Dec 31, 2009 | Ashok Kunjidhapatham | Link Diversity and Load Balancing Across Digital and Optical Express-Thru Nodes |

US20100157847 * | Dec 23, 2008 | Jun 24, 2010 | Dahai Xu | Method and apparatus for providing redundancy for an access network |

US20110051594 * | Mar 3, 2011 | Noblis, Inc. | System and Method for Providing Network Resiliency | |

US20110080846 * | Apr 7, 2011 | Fujitsu Network Communications, Inc. | Determining Paths In A Network With Asymmetric Switches | |

US20110228702 * | Sep 22, 2011 | Noblis, Inc. | System and Method for Providing Network Resiliency | |

US20130163465 * | Dec 21, 2011 | Jun 27, 2013 | Verizon Corporate Services Group Inc. | Method and apparatus for finding diverse physical layer paths in networks |

US20140086065 * | Sep 24, 2012 | Mar 27, 2014 | International Business Machines Corporation | Disjoint multi-paths with service guarantee extension |

US20140099119 * | Oct 7, 2013 | Apr 10, 2014 | Futurewei Technologies, Inc. | Transport Functions Virtualization for Wavelength Division Multiplexing (WDM)-based Optical Networks |

US20140185430 * | Mar 4, 2014 | Jul 3, 2014 | At&T Intellectual Property I, L.P. | System and Method for Restoration in a Multimedia IP Network |

CN102571401A * | Dec 24, 2010 | Jul 11, 2012 | 华为技术有限公司 | Method and device for building backup path and method and device for selecting backup path |

WO2010002915A2 | Jun 30, 2009 | Jan 7, 2010 | Infinera Corporation | Link diversity and load balancing across digital and optical express-thru nodes |

WO2010002915A3 * | Jun 30, 2009 | Apr 8, 2010 | Infinera Corporation | Link diversity and load balancing across digital and optical express-thru nodes |

WO2012083790A1 * | Dec 6, 2011 | Jun 28, 2012 | Huawei Technologies Co., Ltd. | Method and device for establishing backup path, and method and device for selecting backup path |

WO2013067111A1 * | Nov 1, 2012 | May 10, 2013 | Alcatel-Lucent Usa Inc. | An ip fast reroute scheme offering full protection |

Classifications

U.S. Classification | 370/225, 370/238 |

International Classification | H04L12/26 |

Cooperative Classification | H04L45/22, H04L45/123, H04L45/124, H04L45/28, H04L45/00 |

European Classification | H04L45/22, H04L45/124, H04L45/28, H04L45/00, H04L45/123 |

Legal Events

Date | Code | Event | Description |
---|---|---|---|

Jun 30, 2004 | AS | Assignment | Owner name: LUCENT TECHNOLOGIES, INC., NEW JERSEY Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:ALICHERRY, MANSOOR ALI KHAN;SANIEE, IRAJ;SENGUPTA, SUDIPTA;REEL/FRAME:015544/0081 Effective date: 20040521 |

May 26, 2009 | AS | Assignment | Owner name: ALCATEL-LUCENT USA INC., NEW JERSEY Free format text: MERGER;ASSIGNOR:LUCENT TECHNOLOGIES INC.;REEL/FRAME:022730/0955 Effective date: 20081101 |

Jan 10, 2013 | FPAY | Fee payment | Year of fee payment: 4 |

Rotate